Reseach On Hybrid Collaborative Filtering Recommendation Algorithm Based On Weight Redistribution

With the development of Internet technology,the amount of information carried by the Internet has increased dramatically,and the problem of "information overload" has become more and more serious.Search engines are currently unable to meet people's demand for information acquisition,resulting in personalized recommendation system technology.The personalized recommendation system can help users quickly find the information they need from the massive and disorderly information,which alleviates the problem of “information overload” to a certain extent.Collaborative filtering is one of the most widely used and successful technologies in personalized recommendation technology.It has unique advantages in comparing "information overload" with search engines and other recommended technologies.However,with the continuous improvement of user demand and the increase of data size,collaborative filtering technology has also exposed some different problems,among which the cold-start problem is typical.In the collaborative filtering system,the number of historical ratings of users in the system is also greatly different because of different factors such as entering the system time.Traditional collaborative filtering recommends using the same recommendation strategy to generate recommendations for users in the recommendation system,without considering the difference in the number of users' ratings.Based on the above problems,this paper proposes a hybrid collaborative filtering algorithm based on weight redistribution method.The main research contents are as follows:For the system cold-start problem,mitigate the cold-start problem of the collaborative filtering recommendation system by adding the user multi-attribute feature similarity to the similarity calculation of the collaborative filtering algorithm.User multi-attribute similarity is composed of multiple one-dimensional attribute similarities.In the one-dimensional attribute similarity combination method,get the optimal weight distribution of one-dimensional attribute similarity by the analytic hierarchy process(AHP),then get the user's multi-attribute similarity.User multi-attribute similarity does not depend on user behavior data,so there is no coldstart problem,combined with collaborative filtering algorithm,to achieve the purpose of mitigating the system cold-start problem.For the difference in the number of user ratings in the recommendation system,this paper dynamically combines the similarity of user attributes with the similarity of user ratings,and dynamically allocates the number of users in the weight distribution of the two.Dynamically adjust parameter settings based on the number of user ratings In this paper,get the functional relationship between two kinds of similarity weight assignments based on experimental data.After experimental analysis on the public dataset Movielens,the hybrid collaborative filtering algorithm proposed in this paper is better than the traditional collaborative filtering algorithm.